Development of Model-Ant Colony Optimization (ACO) in Route Selection Based on Behavioral User was Transport in Semarang, Indonesia



Similar documents
An ACO Approach to Solve a Variant of TSP

An ant colony optimization for single-machine weighted tardiness scheduling with sequence-dependent setups

Ant Colony Optimization (ACO)

Modified Ant Colony Optimization for Solving Traveling Salesman Problem

An Improved ACO Algorithm for Multicast Routing

Using Ant Colony Optimization for Infrastructure Maintenance Scheduling

STUDY OF PROJECT SCHEDULING AND RESOURCE ALLOCATION USING ANT COLONY OPTIMIZATION 1

INTERNATIONAL JOURNAL OF COMPUTER ENGINEERING & TECHNOLOGY (IJCET)

An Improved Ant Colony Optimization Algorithm for Software Project Planning and Scheduling

A RANDOMIZED LOAD BALANCING ALGORITHM IN GRID USING MAX MIN PSO ALGORITHM

Wireless Sensor Networks Coverage Optimization based on Improved AFSA Algorithm

ACO Hypercube Framework for Solving a University Course Timetabling Problem

MRI Brain Tumor Segmentation Using Improved ACO

A Quantitative Decision Support Framework for Optimal Railway Capacity Planning

Network Load Balancing Using Ant Colony Optimization

Biological inspired algorithm for Storage Area Networks (ACOSAN)

Study on Cloud Computing Resource Scheduling Strategy Based on the Ant Colony Optimization Algorithm

Swarm Intelligence Algorithms Parameter Tuning

Software Project Planning and Resource Allocation Using Ant Colony Optimization with Uncertainty Handling

Web Mining using Artificial Ant Colonies : A Survey

D A T A M I N I N G C L A S S I F I C A T I O N

On-line scheduling algorithm for real-time multiprocessor systems with ACO

Journal of Theoretical and Applied Information Technology 20 th July Vol.77. No JATIT & LLS. All rights reserved.

EA and ACO Algorithms Applied to Optimizing Location of Controllers in Wireless Networks

AN EFFICIENT LOAD BALANCING APPROACH IN CLOUD SERVER USING ANT COLONY OPTIMIZATION

Highway Maintenance Scheduling Using Genetic Algorithm with Microscopic Traffic Simulation

Review of Solving Software Project Scheduling Problem with Ant Colony Optimization

A Proposed Scheme for Software Project Scheduling and Allocation with Event Based Scheduler using Ant Colony Optimization

IAJIT First Online Publication

CLOUD DATABASE ROUTE SCHEDULING USING COMBANATION OF PARTICLE SWARM OPTIMIZATION AND GENETIC ALGORITHM

Ant colony optimization techniques for the vehicle routing problem

HYBRID ACO-IWD OPTIMIZATION ALGORITHM FOR MINIMIZING WEIGHTED FLOWTIME IN CLOUD-BASED PARAMETER SWEEP EXPERIMENTS


VEHICLE ROUTING PROBLEM

Multi-Robot Traffic Planning Using ACO

Multi-Objective Supply Chain Model through an Ant Colony Optimization Approach

vii TABLE OF CONTENTS CHAPTER TITLE PAGE DECLARATION DEDICATION ACKNOWLEDGEMENT ABSTRACT ABSTRAK

PRIORITY SCHEDULING OF URBAN REHABILITATION STREETS TO MINIMIZE CONGESTION, DELAYS, AND ACCIDENTS

International Journal of Emerging Technology & Research

A DISTRIBUTED APPROACH TO ANT COLONY OPTIMIZATION

Finding Liveness Errors with ACO

An Ant Colony Optimization Approach to the Software Release Planning Problem

Evaluation of Test Cases Using ACO and TSP Gulwatanpreet Singh, Sarabjit Kaur, Geetika Mannan CTITR&PTU India

Alpha Cut based Novel Selection for Genetic Algorithm

Proposed Software Testing Using Intelligent techniques (Intelligent Water Drop (IWD) and Ant Colony Optimization Algorithm (ACO))

A Genetic Algorithm Approach for Solving a Flexible Job Shop Scheduling Problem

A Performance Comparison of GA and ACO Applied to TSP

A Comparative Study of Scheduling Algorithms for Real Time Task

A hierarchical multicriteria routing model with traffic splitting for MPLS networks

Obtaining Optimal Software Effort Estimation Data Using Feature Subset Selection

Projects - Neural and Evolutionary Computing

ACO FOR OPTIMAL SENSOR LAYOUT

TEST CASE SELECTION & PRIORITIZATION USING ANT COLONY OPTIMIZATION

Automated SEO. A Market Brew White Paper

COMBINING THE METHODS OF FORECASTING AND DECISION-MAKING TO OPTIMISE THE FINANCIAL PERFORMANCE OF SMALL ENTERPRISES

Network (Tree) Topology Inference Based on Prüfer Sequence

Optimization of Bus Route Planning in Urban Commuter Networks

Distributed Optimization by Ant Colonies

A WEB-BASED TRAFFIC INFORMATION SYSTEM USING WIRELESS COMMUNICATION TECHNIQUES

An ACO algorithm for image compression

Risk Management for IT Security: When Theory Meets Practice

A Binary Model on the Basis of Imperialist Competitive Algorithm in Order to Solve the Problem of Knapsack 1-0

Optimization of ACO for Congested Networks by Adopting Mechanisms of Flock CC

A Survey on Load Balancing Techniques Using ACO Algorithm

On the Interaction and Competition among Internet Service Providers

Design and Analysis of ACO algorithms for edge matching problems

MODELING TRANSPORTATION DEMAND USING EMME/2 A CASE STUDY OF HIROSHIMA URBAN AREA

Solving Traveling Salesman Problem by Using Improved Ant Colony Optimization Algorithm

Performance Analysis of Cloud Computing using Ant Colony Optimization Approach

Comparison of WCA with AODV and WCA with ACO using clustering algorithm

University of British Columbia Co director s(s ) name(s) : John Nelson Student s name

Network Traffic Modelling

How To Understand And Solve A Linear Programming Problem

Cloud database dynamic route scheduling based on polymorphic ant colony optimization algorithm

Evolving an Edge Selection Formula for Ant Colony Optimization

An ACO/VNS Hybrid Approach for a Large-Scale Energy Management Problem

Ohio Standard Small/Medium MPO Travel Demand Forecasting Model. Theoretical Training May 10-11, 2006 Abridged to Freight Focus Sept 2010

Research on the Performance Optimization of Hadoop in Big Data Environment

Effective Load Balancing for Cloud Computing using Hybrid AB Algorithm

Load balancing Static Load Balancing

A Direct Numerical Method for Observability Analysis

Load Balancing in Ad Hoc Networks: Single-path Routing vs. Multi-path Routing

ANT COLONY OPTIMIZATION ALGORITHM FOR RESOURCE LEVELING PROBLEM OF CONSTRUCTION PROJECT

Load Balancing and Termination Detection

Software Framework for Vehicle Routing Problem with Hybrid Metaheuristic Algorithms

DATA AGGREGATION FOR ANT COLONY TECHNIQUES IN WIRELESS SENSOR NETWORK

A hybrid ACO algorithm for the Capacitated Minimum Spanning Tree Problem

Using GIS Network Analyst to Solve a Distribution Center Location Problem in Texas. Texas A&M University, Zachry Department of Civil Engineering

CSE373: Data Structures and Algorithms Lecture 3: Math Review; Algorithm Analysis. Linda Shapiro Winter 2015

Review of Ant Colony Optimization for Software Project Scheduling and Staffing with an Event Based Scheduler

Optimal Planning with ACO

Transcription:

International Refereed Journal of Engineering and Science (IRJES) ISSN (Online) 2319-183X, (Print) 2319-1821 Volume 3, Issue 6 (June 2014), PP.09-18 Development of Model-Ant Colony Optimization (ACO) in Route Selection Based on Behavioral User was Transport in Semarang, Indonesia Siswanto, Joko 1 Jinca M. Yamin 2, Bambang Riyanto 3 Doctoral Student Department of Civil Engineering University of Diponegoro Semarang, Indonesia Professor, Dr.-Ing.,-MSTr.,Ir. in Transportation Engineering, Department of Civil Engineering University of Hasanuddin Makassar, Indonesia Dr.Ir., DEA, in Transportation Engineering, Department of Civil Engineering of Diponegoro, Semarang, Indonesia ABSTRACT: In the selection of transport routes typically use optimization considerations transportation system with aggregative behavior by determining the shortest route or the lowest cost. Determinants of consumer behavior in route selection and decision making dominant (disaggregated). Route selection optimization model based on user behavior can be implemented from the results of model development preference with conjoint analysis. Preferences user behavior seems, is directly proportional to the convenience, the crowd, the facilities, the ease, safety, and inversely proportional to the density. Route selection optimization model with the development of ant - colony optimization substitution formula can be applied with a probability of interaction and preferences as well as the network models incorporate the concept of route selection based on consumer behavior and the physical conditions of the network as the provincial capital of Semarang, Central Java, Indonesia has a lot of road network that connects the origin zone trip up to the trip destination zone, the route selection optimization model based on user behavior can be implemented for the selection of trips in the city of Semarang. Keywords:- Optimization, Consumer Behavior, Systems Network, Route Options. I. INTRODUCTION Selection of land transportation for this route, always using optimization considerations on the network system with aggregative behavior that is based on the determination of the shortest route or the minimum cost. The fact indicates that the consideration of consumer behavior models is very dominant determinant yamg decision. Thus, the selection of design decisions that during the land transport is only considering the shortest route or the lowest cost should consider the concept of consumer behavior as land transportation route selection considerations. Transport route choice models can be developed based on consideration of consumer behavior, the concept of route selection, consideration of psychology and user preference towards the preferred path network. Can be explained that the choice of routes, giving satisfaction to the consumer and the consumer is deemed to have aspects of distance or time and the other is minimal. The design of transport routes, generally based on the optimization of multiple networks to obtain optimum route. Nowadays, highly developed optimization model and a challenge for researchers and practitioners from multiple disciplines, using some optimization techniques that have been developed. Traditional operations research techniques have evolved, but still a lot of shows calculations and complex problem. The result is a metaheuristic search algorithm such as Simulated Annealing (SA), Tabu Search (TS), Genetic Algorithms (GA) and Ant Colony Optimization (ACO) has been applied to obtain good solutions with a good enough computing time. This study uses an algorithm Ant Colony Optimisation (ACO) to solve the unique of route problems, and aims to explore possible solutions to develop a good solution and a small computational time. Flexibility of ACO algorithms using basic probability in the selection of the best route, making the ACO is able to involve several parameters in the selection of the best route. Purpose of this research is the development of transport routes selection model by considering the behavior of the user, using an algorithm Ant Colony Optimisation (ACO), aims to determine consumer preferences based on user behavior is analyzed by conjoint, and determine the model optimization algorithms developed from Ant Colony Optimization (ACO). 9 Page

This study was applied to the road network of the city of Semarang, Central Java, Indonesia with the hope of giving benefits to transport users in obtaining transportation network that connects the origin zone to the destination zone based on user behavior and characteristics of the road network. II. LITERATURE REVIEW 2.1 Theory of Consumer Preferences Heirshleifer and Glazer (1992) suggests that the ideal alternative to the individual preferences of consumer goods can be viewed from two aspects of the law Axiom Law ( revealed preferences), namely : 1. Axiom Comparison, and 2. Axiom of Transitivity. The second axiom when combined will form a proportion of the entire sorting preference that the items are consistently sorted in order of preference somebody, this sorting function is called preference. According to Browning and Zupan (1997) there are three basic assumptions in consumer preferences, namely ; 1. Consumers can rank order of preference in full for all goods in the market. The order of preference indicates the relative degree of pleasure without regard to the price of the goods; 2. Preference is transitivity, this assumption allows people to have a rational and consistent preference; and 3. Consumers will favor the goods in large quantities of the stuff a bit. 2.2 Attribute Theory Approach Attribute approach is based on the assumption that consumers' attention not to the physical product, but rather addressed to the relevant product attributes (Arsyad, 2008). This approach uses analytical tools merged with the indifference curve analysis. Here are all attributes of services resulting from the use or possession of an item. Consumer preferences can be described through indifference curve shows the combinations of two goods that provide the same level of satisfaction to the individual. Level of customer satisfaction will be different (higher / lower) when a combination of the two goods are in a different indifference curve (Nicholson, 2002). Furthermore Nicholson (2002) explains that for arise maximixed utility, the combination of items selected by consumers is limited by budget constraints. This constraint indicates where the combination of these items can be obtained and determine how an individual 's ability to select any combination of categories that provide the highest utility. 2.3 Ant Colony Optimization Ant-based techniques were first proposed by Dorigo et. al (1996) using the ACO to solve the Traveling Salesman Problem (TSP). In ACO, each ant in a herd that runs will leave pheromone (a kind of chemical substances) on the path which it passes. This pheromone, be a signal to fellow ants. Shortest path will leave a stronger signal. The next ant, when deciding which path to choose, tend to choose to follow the path with the strongest signal, so that the shortest path will be found as more ants will pass through the lane. The more ants through a pathway, the stronger the signal in the lane. 2.4 Behavior of Ants An ant k at node i will choose the destination node j in herd next ant with probability equation as follows;...(1) where, α indicates the degree of interest of pheromone and Ni (k) is an alternative option ant k (neighborhood) at the time of being at node i. Neighborhood of ant k in node i will contain all the vertices that can be addressed directly connected to the node i, except node previously visited. 2.5 Addition and Pheromone Evaporation An ant k when passing will leave a pheromone segment. Number pheromone contained in the bypassed segment after ant k ij is given by the following formula;...(2) 10 Page

Increasing the value of pheromone on segment i-j, gives great opportunities of this segment will be selected in the next iteration. After evaluating a number of vertices. Thus, pheromone evaporation will occur according to the following rules;...(3) where ρ (0, 1) is the evaporation rate parameter, A represents a segment or segments that have been passed ant k as part of the trajectory of the nest toward food sources. The decrease in the amount of pheromone enables ants to explore a different path during the search process. This will eliminate the possibility of choosing the path that is less good and help limit the maximum value achieved by a pheromone path. The amount of pheromone is added to the segment i - j by ant k is given as...(4) where Q is a constant and L k is the shortest path through which the ant k. Q value is usually determined by the user, or can be implemented with the following equation: If (i, j) global best path for others...(5) ACO model was initially applied by the inventor Marco Dorigo [Dorigo et al., 1996] to solve the problem Traveling Salesman Problem (TSP). In the TSP, if there are n cities there will be a (γ) segment, and have (γ - 1) ½ a possible route. Distance used in the standard TSP is symmetric distance is the distance between city r to city s equal to the distance between the city s to city r where d (s, r) = d (r, s). The primary role of pheromone evaporation is to prevent stagnation, ie a situation in which all the ants ended up doing the same route. The above process is then repeated until the routes that do reach the maximum number of iterations specified by the user) or does not happen again change solutions. Status transition rules used by Ant system called random proportional rule, which is shown by the following equation; another...(6) where p k (r, s) is the opportunity to move from city to city r s of ant k, τ is the pheromone level, η (r, s) the proximity of the town r and s and β are weights that control the visibility of the pheromone level τ. The first needs to be done is to compute the proximity matrix between city states (visibility), η (r, s), where each entry is 1/distance from r to s. For each ant needs to be done early in the placement where the city will begin the route. For TSP, each ant will start from city 1. From this equation, the segment that is shorter and has a larger amount of pheromone will be greater probability to be selected in the route. The level of pheromone (pheromone trail) change both locally and globally. Global pheromone update is intended to give sections are to be found in the shortest route. After one iteration, all ants complete one route, after all the ants back to the nest will happen with the pheromone updating rules...(7) where ρ (0, 1] is the rate of evaporation or often also called the decay factor and Δτ (k) i, j is the amount of pheromone is added to the segment r-s by the best ant k. goal of the update increases the pheromone is the pheromone value associated with promising path or trajectory that good. artificial ants After completing one route, then it will add the best ant pheromone on sections that have been visited, ie segments are included in the route. while other sections do not change the level of pheromonenya. Amount pheromone Δτ (r, s) is added to each link you visit (r,s) by the best ant is inversely proportional to the total distance of the route, the shorter the higher the amount of pheromone is added to the segment - ruasnya. way is meant to mimic how pheromones These events are accumulated in the real, which is affected by the path length and continuity of the trajectory of live ant time. Global update is formulated as follows; 11 Page

...(8) where (r, s) = (shortest route)-1a or 1/L k, where L k is the shortest path taken by ant k. Local pheromone update rule is useful to prevent the good sections are not selected by all ants. Each time a segment chosen by an ant number pheromonena changed by applying the local update rules:...(9) where τ is a parameter. Local update is also influenced by evaporation as happens in real ants. In other words, the influence of local renewal is to make the level of attractiveness of the existing sections are dynamically changing. Every time an ant uses a segment, then this segment will immediately reduced interest rate (due to the amount of pheromone segment decreased due to evaporation). It will indirectly create another ant will choose the other sections that have not been visited. So the ants will not have a tendency to choose the same path. By creating a different route then there is a higher likelihood of getting one better route than if they all take the same route. 2.6 Model Study Area Modelling for the study area road network and the network structure of the city made with the initial step is the division of the study area into zones and defines the road network. 1. Area Studies The study area is determined as an area which is considered to represent all the information necessary to estimate trip models travel needs in accordance with the objectives of the study. The area bounded by a line cordon study (Figure 2.1). 2. Zones For Small Spatial Analysis Unit Study area is divided into N zones. Each zone is the smallest spatial unit as observation or towing power and movement. Each zone is represented by the center of the zone, as the generation and or pull start of the movement. The main criteria in the zoning in the study area include: a. zone must be able to present as a unit generation and movement or towing b. zone must be smaller than that required to handle traffic on a sub- area study c. zone has a characteristic homogeneity, such as the type of activities, land use and population density of others. d. To facilitate data collection and modeling for the study of transport, zone boundaries usually follow administrative boundaries. The size of zone depends on the size of the study area, to study the transport system of State different size it will be necessary to study the zone unit urban transport. Basically the smaller the broad zone made, the more thorough the research results obtained. Figure 2.1 Regional Model Study (Tamin, O.Z, 2000) 12 Page

Description: Cordon Line Zone Boundary Centroid zones Centroid connector Link Nodes : boundary of the study area : the boundary zone : the central zone : connecting the center of the zone : roads : the point of intersection and the station A road segment (link) is defined as a liaison between the stain with a stain next. III. RESEARCH METHOD In this study, consumer preferences in choosing the route is done by building stimuli that is combination of several attributes which are based on the concept of the transport behavior, namely: 1. The road traffic density 2. Perceptions road comfort 3. Crowds roadside 4. Perceptions on roads facility 5. Ease reaching freight feeder lines from home / office 6. Roads Security Indication - Problem - Purpose and Target - Scope of the Discuss Data Collecting Literature Review Primair Data: Behavior Questionair Sample Behavior Modelling Selection of variables based on behavior Secondary Data: Network Data Elapsed time Network Node Capacity The design is based on Model Preferences Preferences of User Behavior With Conjoint Analysis Variable selection based on network Synthesis of Model Behavior in Network Simulation Analysis Optimization with Ant Colony Optimization Conclusion Selection of Transport Route Figure 3.1. Research Framework 13 Page

3.1 Conjoint Analysis Conjoint analysis is used to determine the relative interests, associated attributes important to customers and utilities they attach to the level or levels of attributes (Supranto, 2004). The steps that must be done in using conjoint analysis is (Supranto, 2004). 1. Formation of the stimulus is done by using the full profile (full - profile procedure), which took six attributes are safe, comfortable, crowded, facilities, easy, and solid. 2. Number of stimuli to be generated is as much as 2 6 = 64 stimuli were obtained from the multiplication between the number of levels for each attribute. 3. Amount of the stimuli is too much that it would confuse the respondents in the sort. To reduce the number of stimuli that is too much, then the reduction is done by using orthogonal array with statistical formula combinations permutations n = 2 6 /2 2 = 16 stimuli. This design assumes that all interactions are not important can be ignored. 4. Stimuli are then used in the questionnaire for sorted by respondents. 5. Subsequently the study subjects were asked to fill in their preference ranking of the 16 combinations of these attributes with values ranking which shows very unpopular highest value and the lowest value indicates the least favored. 6. From the data results to the preference of the 16 stimuli were presented conjoin analysis procedure is then performed with the basic model as follows :...(10) where: μ (X) is the entire utility of an alternative aij is the coefficient of parth-worth or utility of the i-th attribute and level j k i is the number of levels of attribute i m is the number of attributes Xij is a dummy variable i-th attribute level j From the analysis of the coefficients obtained will conjoin parth worth of each attribute level. Part coefficient worth is the value of each attribute level were observed. The order of preference ranking is the sum total of the value of each coefficient of each attribute level in all combinations of attributes, with the following formula; Yb = XP1 + XP2 + Xp3 + Xp4 + XP5 + XP6... (11) where: Yb = preference behavior Xp1 = partworth attribute level 1, XP2 = partworth attribute level 2, Xp3 = partworth attribute level 3, Xp4 = partworth attribute level 4, XP5 = partworth attribute level 5, XP6 = partworth attribute level 6 The relative value factor shows the percentage of each attribute to the preference. To find out the most important attribute for consumers, the interest rate can be calculated attributes. Attribute the difference in the level of interest is the highest and lowest utility. 3.2 Ant Colony Optimisation ACO optimization models modified from the beginning of the Ant System model used by Dorigo et. al (1996). Some modifications were made are as follows: 1. Parameters shortest route The probability of selecting the shortest route is done by using the value of the visibility, which is a measure intended to give effect to each ant in the move towards the next target point with a straight line distance. Because the purpose of ant has been determined, then it is different from the TSP models just getting visibility of the shortest distance alone. The determination of the visibility of a point relative to the target point is as follows: 14 Page

This may point This time point Target point Distance of the current point to the target point Visibility = Distances are likely point to the target point... (12) Values greater visibility will give a greater probability 2. Parameters of behavior perception with 6 preference. The probability of route selection based on the behavior of the six preference secure form, convenient, crowded, facilities, easy, solid values based on the scores obtained from the Conjoint models for each road segment. Value of Conjoint preference scores relative to the maximum score possible. This aspect is called acceptance and value is obtained as follows: Conjoint preference = preference scores 3. Pheromones Parameter Pheromones is a trail left by other ants that previously passed through the segment. Pheromones value greater probability greater route choice. By using all three of these parameters then the acceptance probability is obtained as follows: Prob(i,j,k) = (fer) + (vis) + (pref p) + (1-DS)...(13) Description: (fer) + (vis) + (pref p) + (1 -DS) ) i : index origin node j : index node alternative routes k : index of the target node (end node) : coefficient multiplier pheromones : coefficient multiplier visibility : coefficient multiplier preferences IV. RESULTS AND DISCUSS 4.1 Analysis of Preferences Preference analysis carried out by the results of a survey of 100 respondents transportation users. Conjoint analysis produces utility value (utility) of each level of each attribute and the value of interest (average importance score is how important an attribute to the overall preference. Utility value and the value of the interest generated by the conjoint analysis are shown in the following table: Table 1 Results of conjoint analysis No Atributes Customer Value (%) 1 Comfort (Xp1) 19.581 2 Crowded (Xp2) 13.153 3 Facility (Xp3) 16.947 4 Easy (Xp4) 18.433 5 Safe (Xp5) 15.796 6 Solid (Xp6) 16.090 Table 1 shows that the most important attribute considered is the convenience attribute that to the percentage interest of 19.581%, while most attributes are considered important attribute that is equal to 13.153% crowd. This means that users feeder transport services in the first intermodal service is to pay attention to the comfort of the road into a route. In linear mathematical model of network transport route selection preferences can be obtained as follows: 15 Page

Y b = K 1 Xp 1 + K 2 Xp 2 + K 3 Xp 3 + K 4 Xp 4 + K 5 Xp 5 K 6 Xp 6 Description (Source: Dissertation Unpublish Joko Siswanto 2013): K = 0,260; K = 0,596; K = 0,182; K = 0,154; K = 0,535; K = 0,213 4.2 Optimization Optimization is used to implement a model of perception-preference and network model to determine the best network models. Implementation of route determination in this case is done at Semarang Central Java Indonesia on the road network connecting the O zone to zone E in Figure 4.1 and Figure 4.2 rail and road network data highway with 59 nodes and figure 4.3 optimization results Picture 4.1 Map of Semarang City Picture 4.2 Network Rail Roads and Highways 16 Page

The parameters used in the following optimization: = 2 = 5 = 5 Picture 4.3 Optimization results Rail and Highways Black line shows the preferred route optimization based on intermodal route choice models, as figure 4.3 The preferred route from the starting zone to the node O 1-2-10-11-12-16-24-27-48-44-43-52-57-59 toward the end zone E or other choice of starting zone towards O 4-6-7-8-12-16-24-27-48-44-43-52-57-59 toward the end zone E. V. CONCLUSION AND RECOMMENDATION 5.1 Conclusion Route selection optimization model based on behavior can be applied by first get Conjoint analysis modeling of preference. Preference of the user's behavior is directly proportional to the convenience, the crowd, the facilities, the ease, safety, and inversely proportional to the density. Optimization of route choice models with ant colony optimization can be applied by using the formula of probability of interaction between the preference network model. 5.2 Recommendation This model can be used to the concept of route selection based on consumer behavior and the physical condition of tissue that can be applied in Semarang Central Java Indonesia. 17 Page

REFERENCES [1]. Arsyad, Lincolin, 2008, "Managerial Economics", edition of four, BPFE, Yogyakarta. Central Bureau of Statistics, 2008, "Yogyakarta in Figures". [2]. Browning, Edgar K., and Mark Zupan, 1997, "Microeconomic Theory and Applications", Fifth Edition, Harper Collins College Publishers. [3]. Dorigo, Marco., 1996. The Ant System: Optimization by a colony of cooperating agents IEEE Transactions on Systems, Man, and Cybernetics-Part B, Vol.26, No.1, 1996, pp.1-131 [4]. Dorigo, Marco Thomas Stutzle, 1996. Ant Colony Optimization, A Bradford Book The MIT Press Cambridge, Massachusetts London, England. [5]. Hirshleifer, Jack and Glazer, Amihai, 1992, "Price Theory and Applications", Fifth Edition, Prentice Hall. [6]. Nicholson, Walter, 2002, "Intermediate Microeconomics and Its Application", AlihBahasa, Eighth Edition, grants, Jakarta. [7]. Supranto, J., 2004, "Multivariate Analysis: Meaning and Interpretation", Rineka Copyright, Jakarta. [8]. Tamin, O.Z. 2000, Planning and Modelling Transportation, 2nd ed. ITB, Bandung. Appendix: The algorithm used For each point (i, j) do pheromone = to specify the value end For k = 1 to m do Put Ants (k) at any point at random Determine the value for the distance Lk tempuk Ants (k) and Tk to the set of points visited by ants (k) end ---- repeat ---- For iteration = 1 to iterasi maximum do For k = 1 to m do Calculate point visited Ants (k) by applying the formula of probability (fer) + (vis) + (pref p) + (1-DS) Prob (i, j, k) = (fer) + (vis) + (pref p) + (1-DS)) Greatest value is the probability that a point chosen ants Add ants point visited Ant K to the set T k End for k For k = 1 to m do Calculate the length of each track Lk visited Ants (k) End for k For k = 1 each point (i, j) do Update the pheromone values End for k End for iteration Show route on the ants of the set origin to set goals 18 Page